Repository navigation

fix: initialize hybrid ATOM KV tables and per-request masks - #210

Draft

gspark-etri wants to merge 7 commits into

fix/207-atom-hybrid-runtime

Draft

fix: initialize hybrid ATOM KV tables and per-request masks#210
gspark-etri wants to merge 7 commits into
optimus-rmfrom
fix/207-atom-hybrid-runtime

Commits on Oct 23, 2026
- fix: KV-cache generation at batch size >= 2 (#164) (#187)
  
  gspark-etri
  and
  claude
  committed
  7455ba1
- fix: per-rank checkpoint names for TP/DP (#165) (#188)
  
  gspark-etri
  and
  claude
  committed
  3d07128
- fix: scope MPS daemon detection and cleanup to our pipe dir (#166) (#189)
  
  gspark-etri
  and
  claude
  committed
  0e441d2
- Merge pull request #190 from ai-computing/fix/164-hotfix-v4.1.2
  ememos
  authored
  fd89265
- Merge engine prerequisite for hybrid cache runtime fix
  gspark-etri
  committed
  3eaa897
- fix: initialize hybrid ATOM cache tables and request masks (#207)
  gspark-etri
  committed
  6b5d8f3
- test: attach measured hybrid runtime outputs
  gspark-etri
  committed
  21f2c87